HOMICIDES - County Level

## # A tibble: 2 x 4
##   Higher Lower `No Diff` total
##    <dbl> <dbl>     <dbl> <dbl>
## 1   17    18        19      54
## 2   31.5  33.3      35.2   100




SUICIDES - County Level

## # A tibble: 2 x 4
##   Higher Lower `No Diff` total
##    <dbl> <dbl>     <dbl> <dbl>
## 1   27    5         26      58
## 2   46.6  8.62      44.8   100




HOMICIDES - Community Level

## # A tibble: 2 x 4
##   Higher Lower `No Diff` total
##    <dbl> <dbl>     <dbl> <dbl>
## 1   73   209       209     491
## 2   14.9  42.6      42.6   100




SUICIDES - Community Level

## # A tibble: 2 x 4
##   Higher Lower `No Diff` total
##    <dbl> <dbl>     <dbl> <dbl>
## 1  101   148       282     531
## 2   19.0  27.9      53.1   100




HOMICIDES – INCREASES ONLY




SDOH Work

quick “exploration” of distributions of the three (plus) varibles we are “correlating”

histograms to look at the distribution of each

  • community poverty and education from 2017 5-year ACS data, so covers 2013-2017; education is for population 25 and older
  • community homicide is age-adjusted rate for 2013-2017 data combined, 39 (of 561) communities missing data, mostly do to cell suppression

observations:

  • poverty - normal-ish, slight right skew (can’t be smaller than 0 or larger than 100)
  • education - skewed left
  • homicide - strong skew right



now looking at the “raw” associations

  • appears to have strong linear association, but “spread” increasing with poverty (transform?)

  • non-linear relationship - log-linear (i.e. exponential) explore transformation;
  • explore difference in pattern between poverty and education with homicide


check that this is not a coding error with eduction by comapring “our” ACS data pull to HPI

  • not a coding error (lack of exact correlation is from different years of data)

“added value” exploration with:

  • linear modeling lines
  • color based on CHANGES in homicide rates
  • dot size based on number of homicides
  • very messy pop-up to get county/community name